In order to meet the growing demand for internet bandwidth with traffic growth rates around 40-50% per year, telecommunication component providers face the task of increasing the spectral efficiency of modulation formats for fiber transmission. After 10 Gbit/s systems became successful in the 1990's, solutions for 40Gbit/s became available in the last years. Standardization and research are now focused on the development of 100 Gbit/s systems. Coherent polarization multiplex systems with quadrature phase shift keying QPSK or differential quadrature phase shift keying (DQPSK) are the most likely modulation format for next generation systems. Since polarization multiplexing utilizes orthogonal light polarizations, it is possible to transmit a signal at a rate of ˜25-28 Gigasymbols per second, thus fitting into the standard 50 GHz grid for DWDM optical systems. Coherent signal reception makes it possible to compensate linear transmission impairments like chromatic dispersion and polarization-mode dispersion after sampling in the digital domain. Here research and development faces the challenge of digital signal processing algorithms and chip design.
FIG. 1 shows an exemplary coherent receiver for polarization multiplex signals. A received signal comprising two orthogonal optical signals is split by a polarisation beam splitter 1 into two orthogonal component signals x and y. Each of these component signals is split by optical 90°—hybrids 2 and 3 into an in-phase component xi; yi and a quadrature-phase component xq; yq. Therefore frequency and phase of a local carrier generated by a local oscillator 4 must be adjusted by a carrier recovery unit 12 to agree with that of the received polarisation multiplex signal.
After analogue-to-digital conversion by AD-converters (ADC) 5-8 a sampled and quantized representation of the received optical signal is available in digital form referred to as component values XI, XQ; YI, YQ. Such values contain statistic noisy distortions, deterministic channel degradations such as chromatic dispersion, and random time-varying distortions mainly due to polarization effects. A dispersion compensation unit 9 is usually added for first coarse chromatic dispersion compensation.
In addition, a clock recovery subsystem 10 is necessary extracting a correct sampling clock frequency and a correct sampling clock phase from the received signal. In the literature, several approaches to timing information extraction have been proposed for digital signals, in particular:
F. M. Gardner describes “A BPSK/QPSK Timing-Error Detector for Sampled Receivers”, IEEE Transactions on Communications, Vol. COM-34, No. 5, May 1986, pp. 423-429, and M. Oerder and H. Meyer describe a “Digital filter and square timing recovery,” IEEE. Trans. Comm., vol. 36, pp. 605-612, May 1988. Both phase error detectors are fed with a single optical transmission signal.
The polarization of the incoming optical polarisation multiplex signal varies unpredictably over time and it is thus randomly misaligned with respect to the reference axes of the polarization beam splitter 1 used at the receiver's input to separate the incoming polarization multiplexed signal components. This causes the orthogonal optical signals to mix (polarization mixing) into a linear combination dependent on a polarization mixing angle α between the incoming signal's polarizations and the reference axes of the polarization beam splitter. Furthermore, the received orthogonal optical signals experienced a random relative delay due to differential group delay (DGD) effects, e.g. according to polarization mode dispersion. As a result, also the derived electrical signal represented by digital values consists of a random linear combination of the transmitted orthogonal signals additionally affected by a random phase misalignment.
The conventional phase error detectors described by F. M. Gardner or M. Oerder can be used to adjust sampling frequency and phase in a phase locked loop (PLL). These phase detectors assume an already fully equalized input signal, where the input polarization components are phase-aligned and the QPSK components (I and Q) are perfectly separated and not an arbitrary linear combination of the orthogonal component signals x and y.
The Gardner phase error detector's output signal as a function of the phase error possesses a horizontal sinusoidal shape and is commonly termed s-curve. Its amplitude or its maximum derivation is termed by Gardner as “phase detector gain factor” indicating the performance quality. This “phase detector gain factor” is here referred to as “gain coefficient”. In presence of a and DGD effects the phase error information provided by these algorithms degrades significantly according to input signal conditions. This is illustrated in FIG. 2 where a normalized gain coefficient K/KREF (KREF—gain coefficient referent value) of the Gardner phase error detector is plotted versus the phase difference DGD/T (DGD—differential group delay; T—symbol duration) between orthogonal polarisation signals (termed y1 and yQ by Gardner) and for several values of the polarization mixing angle a. FIG. 2 clearly shows that in the worst case for a=n/4 and DGD=T/2, the phase information contained in the original orthogonal optical signals adds destructively and the normalized gain coefficient K/KREF vanishes, leaving a PLL without any valid control information, rendering the loop inoperable.
Following the clock recovery subsystem, the receiver comprises also a butterfly equalizer 11 reconstructing the original orthogonal signals and compensating distortions. The Regained symbol values D1(n), D2(n) are then fed to a carrier recovery unit 12 correcting frequency and phase mismatches between input signal's and local oscillator's carriers. At the output of the carrier recovery unit, the QPSK signal constellation is constant and correctly positioned on the complex (I/Q) plane. The symbols D1(n), D2(n) are fed to a symbol estimation (decoding) unit 13 which outputs regained data signals DS1, DS2. These signals are then fed to a parallel-serial-converter 14 and converted into a serial data signal SDS.